------------------------------------------------------------------------------- name: log: Q:\C-modelling\runmlwin\website\logfiles\2020-03-27\16\16_Multiple > _Membership_Models.smcl log type: smcl opened on: 27 Mar 2020, 18:00:34 . **************************************************************************** . * MLwiN MCMC Manual . * . * 16 Multiple Membership Models . . . . . . . . . . . . . . . . . . . . 231 . * . * Browne, W. J. (2009). MCMC Estimation in MLwiN, v2.26. Centre for . * Multilevel Modelling, University of Bristol. . **************************************************************************** . * Stata do-file to replicate all analyses using runmlwin . * . * George Leckie and Chris Charlton, . * Centre for Multilevel Modelling, 2012 . * http://www.bristol.ac.uk/cmm/software/runmlwin/ . **************************************************************************** . . * 16.1 Notation and weightings . . . . . . . . . . . . . . . . . . . . . 232 . . * 16.2 Office workers salary dataset . . . . . . . . . . . . . . . . . . 232 . . use "http://www.bristol.ac.uk/cmm/media/runmlwin/wage1.dta", clear . . describe Contains data from http://www.bristol.ac.uk/cmm/media/runmlwin/wage1.dta obs: 3,022 vars: 21 21 Oct 2011 12:19 ------------------------------------------------------------------------------- storage display value variable name type format label variable label ------------------------------------------------------------------------------- id int %9.0g company int %9.0g company2 int %9.0g company3 int %9.0g company4 int %9.0g age byte %9.0g parttime byte %9.0g sex byte %9.0g cons byte %9.0g earnings float %9.0g logearn float %9.0g numjobs byte %9.0g weight1 float %9.0g weight2 float %9.0g weight3 float %9.0g weight4 float %9.0g ew1 float %9.0g ew2 float %9.0g ew3 float %9.0g ew4 float %9.0g age_40 byte %9.0g ------------------------------------------------------------------------------- Sorted by: . . histogram earnings (bin=34, start=2.4000001, width=3.9176472) . . histogram logearn (bin=34, start=.87546879, width=.11865414) . . . . * 16.3 Models for the earnings data . . . . . . . . . . . . . . . . . . .235 . . quietly runmlwin logearn cons age_40 numjobs, /// > level2(company:) /// > level1(id: cons) /// > nopause . . runmlwin logearn cons age_40 numjobs, /// > level2(company:) /// > level1(id: cons) /// > mcmc(on) initsprevious /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (hierarchical) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 1.59 Deviance (dbar) = 5199.58 Deviance (thetabar) = 5195.62 Effective no. of pars (pd) = 3.96 Bayesian DIC = 5203.54 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.079718 .0301277 5113 0.000 3.020516 3.139162 age_40 | .0121912 .0010211 5008 0.000 .0101969 .0142025 numjobs | -.1296681 .0237898 4955 0.000 -.1757482 -.0823093 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .3272223 .0084424 4551 .3112467 .3443396 ------------------------------------------------------------------------------ . . quietly runmlwin logearn cons age_40 numjobs sex parttime, /// > level2(company:) /// > level1(id: cons) /// > nopause . . runmlwin logearn cons age_40 numjobs sex parttime, /// > level2(company:) /// > level1(id: cons) /// > mcmc(on) initsprevious /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (hierarchical) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 1.88 Deviance (dbar) = 4989.99 Deviance (thetabar) = 4983.98 Effective no. of pars (pd) = 6.02 Bayesian DIC = 4996.01 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.085226 .0300767 5444 0.000 3.026191 3.142926 age_40 | .0108051 .0010154 4589 0.000 .008826 .0127701 numjobs | -.0329378 .0240924 5469 0.088 -.0801208 .0151424 sex | -.2115666 .0207766 5547 0.000 -.2518489 -.1707307 parttime | -.3891148 .0366846 5255 0.000 -.4604206 -.3178887 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .3054343 .007966 4980 .2904407 .3217833 ------------------------------------------------------------------------------ . . correlate parttime sex numjobs (obs=3,022) | parttime sex numjobs -------------+--------------------------- parttime | 1.0000 sex | 0.0399 1.0000 numjobs | 0.2908 0.1014 1.0000 . . . . . * 16.4 Fitting multiple membership models to the dataset . . . . . . . . 237 . . tabulate numjobs numjobs | Freq. Percent Cum. ------------+----------------------------------- 1 | 2,496 82.59 82.59 2 | 472 15.62 98.21 3 | 52 1.72 99.93 4 | 2 0.07 100.00 ------------+----------------------------------- Total | 3,022 100.00 . . quietly runmlwin logearn cons age_40 sex parttime, /// > level2(company: cons) /// > level1(id: cons) /// > nopause . . matrix b = e(b) . . matrix V = e(V) . . runmlwin logearn cons age_40 sex parttime, /// > level2(company: cons) /// > level1(id: cons) /// > mcmc(on) initsb(b) initsv(V) /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (hierarchical) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 3.22 Deviance (dbar) = 4422.20 Deviance (thetabar) = 4312.28 Effective no. of pars (pd) = 109.92 Bayesian DIC = 4532.12 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.039154 .0234328 596 0.000 2.993151 3.084558 age_40 | .0113165 .0009351 4464 0.000 .0095258 .0131169 sex | -.2119524 .0191381 4945 0.000 -.2485963 -.1732622 parttime | -.3986974 .0328679 4481 0.000 -.4661018 -.3339319 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 2: company | var(cons) | .0511896 .0078875 2034 .037527 .0683178 -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .2530676 .0066014 4338 .2404853 .2660507 ------------------------------------------------------------------------------ . . . runmlwin logearn cons age_40 sex parttime, /// > level2(company: cons, mmids(company-company4) mmweights(weight1-weigh > t4) residuals(u)) /// > level1(id: cons) /// > mcmc(on) initsb(b) initsv(V) /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (cross-classified) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 3.38 Deviance (dbar) = 4355.00 Deviance (thetabar) = 4240.98 Effective no. of pars (pd) = 114.02 Bayesian DIC = 4469.03 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.04037 .0243765 520 0.000 2.992612 3.087447 age_40 | .0113843 .0009217 4568 0.000 .0096111 .0131587 sex | -.2169319 .0189792 4901 0.000 -.2533067 -.1785266 parttime | -.4036873 .03241 4710 0.000 -.4699596 -.3396066 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 2: company | var(cons) | .0585735 .0086528 2289 .0437406 .0774 -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .2474913 .0064388 4358 .2352251 .2601698 ------------------------------------------------------------------------------ . . . . * 16.5 Residuals in multiple membership models . . . . . . . . . . . . . 240 . . preserve . . rename company company1 . . keep company? id u0_? u0se_? . . reshape long company u0_ u0se_, i(id) j(order) (note: j = 1 2 3 4) Data wide -> long ----------------------------------------------------------------------------- Number of obs. 3022 -> 12088 Number of variables 13 -> 5 j variable (4 values) -> order xij variables: company1 company2 ... company4 -> company u0_1 u0_2 ... u0_4 -> u0_ u0se_1 u0se_2 ... u0se_4 -> u0se_ ----------------------------------------------------------------------------- . . drop id order . . rename u0_ u0 . . rename u0se_ u0se . . drop if u0==. (8,484 observations deleted) . . duplicates drop Duplicates in terms of all variables (3,463 observations deleted) . . egen u0rank = rank(u0) . . serrbar u0 u0se u0rank, yline(0) scale(1.4) . . restore . . gen companyno54 = (company==54) + (company2==54) + (company3==54) + (company4 > ==54) . . gen companyno67 = (company==67) + (company2==67) + (company3==67) + (company4 > ==67) . . quietly runmlwin logearn cons age_40 sex parttime companyno54 companyno67, // > / > level2(company: cons) /// > level1(id: cons) /// > nopause . . matrix b = e(b) . . matrix V = e(V) . . runmlwin logearn cons age_40 sex parttime companyno54 companyno67, /// > level2(company: cons, mmids(company-company4) mmweights(weight1-weigh > t4)) /// > level1(id: cons) /// > mcmc(on) initsb(b) initsv(V) /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (cross-classified) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 3.63 Deviance (dbar) = 4356.84 Deviance (thetabar) = 4249.25 Effective no. of pars (pd) = 107.59 Bayesian DIC = 4464.43 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.027504 .0223699 748 0.000 2.984929 3.071763 age_40 | .0113756 .0009335 5056 0.000 .0095689 .013247 sex | -.2176152 .0188775 5024 0.000 -.2558853 -.1812743 parttime | -.4112686 .0322735 4805 0.000 -.4741918 -.3480687 companyno54 | .7570794 .188503 627 0.000 .3878597 1.129479 companyno67 | .8696657 .2095539 524 0.000 .4635231 1.300882 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 2: company | var(cons) | .0443871 .0071716 1897 .0322091 .0603338 -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .2476207 .0065802 4159 .2350137 .2609146 ------------------------------------------------------------------------------ . . . . * 16.6 Alternative weights for multiple membership models . . . . . . . .243 . . runmlwin logearn cons age_40 sex parttime companyno54 companyno67, /// > level2(company: cons, mmids(company-company4) mmweights(ew1-ew4)) /// > level1(id: cons) /// > mcmc(on) initsb(b) initsv(V) /// > nopause MLwiN 3.05 multilevel model Number of obs = 3022 Normal response model (cross-classified) Estimation algorithm: MCMC ----------------------------------------------------------- | No. of Observations per Group Level Variable | Groups Minimum Average Maximum ----------------+------------------------------------------ company | 141 2 21.4 49 ----------------------------------------------------------- Burnin = 500 Chain = 5000 Thinning = 1 Run time (seconds) = 3.56 Deviance (dbar) = 4369.45 Deviance (thetabar) = 4262.49 Effective no. of pars (pd) = 106.96 Bayesian DIC = 4476.41 ------------------------------------------------------------------------------ logearn | Mean Std. Dev. ESS P [95% Cred. Interval] -------------+---------------------------------------------------------------- cons | 3.026143 .0224423 752 0.000 2.983602 3.070602 age_40 | .0113654 .0009356 5070 0.000 .0095601 .0132456 sex | -.2165363 .0189102 5022 0.000 -.254659 -.1803057 parttime | -.4108328 .0323028 4836 0.000 -.4738127 -.3475643 companyno54 | .7840548 .1944889 586 0.000 .4035939 1.172297 companyno67 | .868183 .2122001 514 0.000 .4579902 1.305356 ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int] -----------------------------+------------------------------------------------ Level 2: company | var(cons) | .0446553 .0072482 1868 .0323424 .0608277 -----------------------------+------------------------------------------------ Level 1: id | var(cons) | .2486563 .0066082 4151 .2359847 .2619662 ------------------------------------------------------------------------------ . . . . * 16.7 Multiple membership multiple classification (MMMC) models . . . . 244 . . * Chapter learning outcomes . . . . . . . . . . . . . . . . . . . . . . .245 . . . . . . **************************************************************************** . exit end of do-file